Machine learning and the Schrödinger equation

Abstract

Abstract In this research several methods of machine learning (ML), such as decision trees and linear regression were used to predict ground-state energy (GSE) of an electron in a potential well. Analysis has been done for various types of the potentials: ones with exact solution as well as ones with only numerical one. It was shown that some methods can map the analytical solution and predict GSE with chemical accuracy along with ability to find GSE for systems which does not have analytical solutions. To increase the accuracy and performance of the ML algorithms we proposed several methods of data preprocessing.